Switch to conda-forge ocl-icd loader packages

[antonwolfy]

Replaces the intel-gpu-ocl-icd-system run dependency with the equivalent OpenCL ICD loader packages from conda-forge:

• Linux: ocl-icd-system
• Windows: khronos-opencl-icd-loader

The -system variant on Linux is intentional: it symlinks the system-installed OpenCL ICD loader into the environment rather than bundling its own, so the Intel GPU driver's registered ICD is discovered correctly. Using the bundling ocl-icd package instead shadows the system loader and breaks device enumeration.

The # [linux] / # [win] selectors are required because the two packages are published for disjoint platforms on conda-forge (ocl-icd-system is Linux-only; khronos-opencl-icd-loader is Windows/macOS-only). dpnp builds only on Linux and
Windows, so no macOS branch is needed.

• Have you provided a meaningful PR description?
• Have you added a test, reproducer or referred to an issue with a reproducer?
• Have you tested your changes locally for CPU and GPU devices?
• Have you made sure that new changes do not introduce compiler warnings?
• Have you checked performance impact of proposed changes?
• Have you added documentation for your changes, if necessary?
• Have you added your changes to the changelog?

[github-actions]

View rendered docs @ https://intelpython.github.io/dpnp/pull/2905/index.html

[github-actions]

Array API standard conformance tests for dpnp=0.21.0dev3=py314h509198e_24 ran successfully.
Passed: 1376
Failed: 1
Skipped: 5

[coveralls]

coverage: 78.411%. remained the same — switch-to-conda-forge-ocl-icd into master